Forum Discussion
@Morgaine I think that the slow ramp up is just an artifact of the advert based sampling method that APNIC use. We saw this in the retrospective talks from the Sky team where there was a distinct lag in the numbers reported by APNIC vs the more direct measurements from the likes of Akamai and the Internet Society. I would suggest that they switched on this trial is two batches, one small (1000 users?) then the rest (9000ish users?). If they had all the participants and infrastructure lined up and ready making a new trial live wouldn't have been the same level of work once they were ready to start. Of course getting to the "ready to start" point is where all the work goes...
@davefiddes:
I'm not convinced that there is a significant lag in APNIC's readings when averaged over a day.
We have evidence of their close-to-instantaneous response on the graph at the 2nd red bar (sudden rise at 2017-12-14) and 5th red bar (sudden fall at 2018-04-10) -- at this x-resolution they are both near-vertical lines. When an IPv6-enabled population of people already exists, APNIC seems to detect the change in the number of addresses used very promptly when the IPv6 provisioning is switched on or off. And conversely, looking at the slow and fairly linear fall in IPv6 activity before the 3rd red bar (2018-01-31) and the similar slow rise in counts after it, it's clear that some kind of operations within VM can produce slow linear changes in the measurements too.
It's very hard to attribute causes to these observations with confidence, but I think we can definitely say that there are two distinct types of change in IPv6 activity observable by APNIC and actually being observed at VM.
Related Content
- 6 months ago
- 8 months ago
- 8 months ago